Loading...
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 | /* * Copyright (c) Yann Collet, Facebook, Inc. * All rights reserved. * * This source code is licensed under both the BSD-style license (found in the * LICENSE file in the root directory of this source tree) and the GPLv2 (found * in the COPYING file in the root directory of this source tree). * You may select, at your option, one of the above-listed licenses. */ /* zstd_decompress_block : * this module takes care of decompressing _compressed_ block */ /*-******************************************************* * Dependencies *********************************************************/ #include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */ #include "../common/compiler.h" /* prefetch */ #include "../common/cpu.h" /* bmi2 */ #include "../common/mem.h" /* low level memory routines */ #define FSE_STATIC_LINKING_ONLY #include "../common/fse.h" #define HUF_STATIC_LINKING_ONLY #include "../common/huf.h" #include "../common/zstd_internal.h" #include "zstd_decompress_internal.h" /* ZSTD_DCtx */ #include "zstd_ddict.h" /* ZSTD_DDictDictContent */ #include "zstd_decompress_block.h" /*_******************************************************* * Macros **********************************************************/ /* These two optional macros force the use one way or another of the two * ZSTD_decompressSequences implementations. You can't force in both directions * at the same time. */ #if defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \ defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG) #error "Cannot force the use of the short and the long ZSTD_decompressSequences variants!" #endif /*_******************************************************* * Memory operations **********************************************************/ static void ZSTD_copy4(void* dst, const void* src) { ZSTD_memcpy(dst, src, 4); } /*-************************************************************* * Block decoding ***************************************************************/ /*! ZSTD_getcBlockSize() : * Provides the size of compressed block from block header `src` */ size_t ZSTD_getcBlockSize(const void* src, size_t srcSize, blockProperties_t* bpPtr) { RETURN_ERROR_IF(srcSize < ZSTD_blockHeaderSize, srcSize_wrong, ""); { U32 const cBlockHeader = MEM_readLE24(src); U32 const cSize = cBlockHeader >> 3; bpPtr->lastBlock = cBlockHeader & 1; bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3); bpPtr->origSize = cSize; /* only useful for RLE */ if (bpPtr->blockType == bt_rle) return 1; RETURN_ERROR_IF(bpPtr->blockType == bt_reserved, corruption_detected, ""); return cSize; } } /* Allocate buffer for literals, either overlapping current dst, or split between dst and litExtraBuffer, or stored entirely within litExtraBuffer */ static void ZSTD_allocateLiteralsBuffer(ZSTD_DCtx* dctx, void* const dst, const size_t dstCapacity, const size_t litSize, const streaming_operation streaming, const size_t expectedWriteSize, const unsigned splitImmediately) { if (streaming == not_streaming && dstCapacity > ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH + litSize + WILDCOPY_OVERLENGTH) { /* room for litbuffer to fit without read faulting */ dctx->litBuffer = (BYTE*)dst + ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH; dctx->litBufferEnd = dctx->litBuffer + litSize; dctx->litBufferLocation = ZSTD_in_dst; } else if (litSize > ZSTD_LITBUFFEREXTRASIZE) { /* won't fit in litExtraBuffer, so it will be split between end of dst and extra buffer */ if (splitImmediately) { /* won't fit in litExtraBuffer, so it will be split between end of dst and extra buffer */ dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH; dctx->litBufferEnd = dctx->litBuffer + litSize - ZSTD_LITBUFFEREXTRASIZE; } else { /* initially this will be stored entirely in dst during huffman decoding, it will partially shifted to litExtraBuffer after */ dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize; dctx->litBufferEnd = (BYTE*)dst + expectedWriteSize; } dctx->litBufferLocation = ZSTD_split; } else { /* fits entirely within litExtraBuffer, so no split is necessary */ dctx->litBuffer = dctx->litExtraBuffer; dctx->litBufferEnd = dctx->litBuffer + litSize; dctx->litBufferLocation = ZSTD_not_in_dst; } } /* Hidden declaration for fullbench */ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx, const void* src, size_t srcSize, void* dst, size_t dstCapacity, const streaming_operation streaming); /*! ZSTD_decodeLiteralsBlock() : * Where it is possible to do so without being stomped by the output during decompression, the literals block will be stored * in the dstBuffer. If there is room to do so, it will be stored in full in the excess dst space after where the current * block will be output. Otherwise it will be stored at the end of the current dst blockspace, with a small portion being * stored in dctx->litExtraBuffer to help keep it "ahead" of the current output write. * * @return : nb of bytes read from src (< srcSize ) * note : symbol not declared but exposed for fullbench */ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx, const void* src, size_t srcSize, /* note : srcSize < BLOCKSIZE */ void* dst, size_t dstCapacity, const streaming_operation streaming) { DEBUGLOG(5, "ZSTD_decodeLiteralsBlock"); RETURN_ERROR_IF(srcSize < MIN_CBLOCK_SIZE, corruption_detected, ""); { const BYTE* const istart = (const BYTE*) src; symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3); switch(litEncType) { case set_repeat: DEBUGLOG(5, "set_repeat flag : re-using stats from previous compressed literals block"); RETURN_ERROR_IF(dctx->litEntropy==0, dictionary_corrupted, ""); ZSTD_FALLTHROUGH; case set_compressed: RETURN_ERROR_IF(srcSize < 5, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3"); { size_t lhSize, litSize, litCSize; U32 singleStream=0; U32 const lhlCode = (istart[0] >> 2) & 3; U32 const lhc = MEM_readLE32(istart); size_t hufSuccess; size_t expectedWriteSize = MIN(ZSTD_BLOCKSIZE_MAX, dstCapacity); switch(lhlCode) { case 0: case 1: default: /* note : default is impossible, since lhlCode into [0..3] */ /* 2 - 2 - 10 - 10 */ singleStream = !lhlCode; lhSize = 3; litSize = (lhc >> 4) & 0x3FF; litCSize = (lhc >> 14) & 0x3FF; break; case 2: /* 2 - 2 - 14 - 14 */ lhSize = 4; litSize = (lhc >> 4) & 0x3FFF; litCSize = lhc >> 18; break; case 3: /* 2 - 2 - 18 - 18 */ lhSize = 5; litSize = (lhc >> 4) & 0x3FFFF; litCSize = (lhc >> 22) + ((size_t)istart[4] << 10); break; } RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled"); RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected, ""); RETURN_ERROR_IF(litCSize + lhSize > srcSize, corruption_detected, ""); RETURN_ERROR_IF(expectedWriteSize < litSize , dstSize_tooSmall, ""); ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 0); /* prefetch huffman table if cold */ if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) { PREFETCH_AREA(dctx->HUFptr, sizeof(dctx->entropy.hufTable)); } if (litEncType==set_repeat) { if (singleStream) { hufSuccess = HUF_decompress1X_usingDTable_bmi2( dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, ZSTD_DCtx_get_bmi2(dctx)); } else { hufSuccess = HUF_decompress4X_usingDTable_bmi2( dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, ZSTD_DCtx_get_bmi2(dctx)); } } else { if (singleStream) { #if defined(HUF_FORCE_DECOMPRESS_X2) hufSuccess = HUF_decompress1X_DCtx_wksp( dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->workspace, sizeof(dctx->workspace)); #else hufSuccess = HUF_decompress1X1_DCtx_wksp_bmi2( dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->workspace, sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx)); #endif } else { hufSuccess = HUF_decompress4X_hufOnly_wksp_bmi2( dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->workspace, sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx)); } } if (dctx->litBufferLocation == ZSTD_split) { ZSTD_memcpy(dctx->litExtraBuffer, dctx->litBufferEnd - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE); ZSTD_memmove(dctx->litBuffer + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH, dctx->litBuffer, litSize - ZSTD_LITBUFFEREXTRASIZE); dctx->litBuffer += ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH; dctx->litBufferEnd -= WILDCOPY_OVERLENGTH; } RETURN_ERROR_IF(HUF_isError(hufSuccess), corruption_detected, ""); dctx->litPtr = dctx->litBuffer; dctx->litSize = litSize; dctx->litEntropy = 1; if (litEncType==set_compressed) dctx->HUFptr = dctx->entropy.hufTable; return litCSize + lhSize; } case set_basic: { size_t litSize, lhSize; U32 const lhlCode = ((istart[0]) >> 2) & 3; size_t expectedWriteSize = MIN(ZSTD_BLOCKSIZE_MAX, dstCapacity); switch(lhlCode) { case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */ lhSize = 1; litSize = istart[0] >> 3; break; case 1: lhSize = 2; litSize = MEM_readLE16(istart) >> 4; break; case 3: lhSize = 3; litSize = MEM_readLE24(istart) >> 4; break; } RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled"); RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, ""); ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1); if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) { /* risk reading beyond src buffer with wildcopy */ RETURN_ERROR_IF(litSize+lhSize > srcSize, corruption_detected, ""); if (dctx->litBufferLocation == ZSTD_split) { ZSTD_memcpy(dctx->litBuffer, istart + lhSize, litSize - ZSTD_LITBUFFEREXTRASIZE); ZSTD_memcpy(dctx->litExtraBuffer, istart + lhSize + litSize - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE); } else { ZSTD_memcpy(dctx->litBuffer, istart + lhSize, litSize); } dctx->litPtr = dctx->litBuffer; dctx->litSize = litSize; return lhSize+litSize; } /* direct reference into compressed stream */ dctx->litPtr = istart+lhSize; dctx->litSize = litSize; dctx->litBufferEnd = dctx->litPtr + litSize; dctx->litBufferLocation = ZSTD_not_in_dst; return lhSize+litSize; } case set_rle: { U32 const lhlCode = ((istart[0]) >> 2) & 3; size_t litSize, lhSize; size_t expectedWriteSize = MIN(ZSTD_BLOCKSIZE_MAX, dstCapacity); switch(lhlCode) { case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */ lhSize = 1; litSize = istart[0] >> 3; break; case 1: lhSize = 2; litSize = MEM_readLE16(istart) >> 4; break; case 3: lhSize = 3; litSize = MEM_readLE24(istart) >> 4; RETURN_ERROR_IF(srcSize<4, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4"); break; } RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled"); RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected, ""); RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, ""); ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1); if (dctx->litBufferLocation == ZSTD_split) { ZSTD_memset(dctx->litBuffer, istart[lhSize], litSize - ZSTD_LITBUFFEREXTRASIZE); ZSTD_memset(dctx->litExtraBuffer, istart[lhSize], ZSTD_LITBUFFEREXTRASIZE); } else { ZSTD_memset(dctx->litBuffer, istart[lhSize], litSize); } dctx->litPtr = dctx->litBuffer; dctx->litSize = litSize; return lhSize+1; } default: RETURN_ERROR(corruption_detected, "impossible"); } } } /* Default FSE distribution tables. * These are pre-calculated FSE decoding tables using default distributions as defined in specification : * https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#default-distributions * They were generated programmatically with following method : * - start from default distributions, present in /lib/common/zstd_internal.h * - generate tables normally, using ZSTD_buildFSETable() * - printout the content of tables * - pretify output, report below, test with fuzzer to ensure it's correct */ /* Default FSE distribution table for Literal Lengths */ static const ZSTD_seqSymbol LL_defaultDTable[(1<<LL_DEFAULTNORMLOG)+1] = { { 1, 1, 1, LL_DEFAULTNORMLOG}, /* header : fastMode, tableLog */ /* nextState, nbAddBits, nbBits, baseVal */ { 0, 0, 4, 0}, { 16, 0, 4, 0}, { 32, 0, 5, 1}, { 0, 0, 5, 3}, { 0, 0, 5, 4}, { 0, 0, 5, 6}, { 0, 0, 5, 7}, { 0, 0, 5, 9}, { 0, 0, 5, 10}, { 0, 0, 5, 12}, { 0, 0, 6, 14}, { 0, 1, 5, 16}, { 0, 1, 5, 20}, { 0, 1, 5, 22}, { 0, 2, 5, 28}, { 0, 3, 5, 32}, { 0, 4, 5, 48}, { 32, 6, 5, 64}, { 0, 7, 5, 128}, { 0, 8, 6, 256}, { 0, 10, 6, 1024}, { 0, 12, 6, 4096}, { 32, 0, 4, 0}, { 0, 0, 4, 1}, { 0, 0, 5, 2}, { 32, 0, 5, 4}, { 0, 0, 5, 5}, { 32, 0, 5, 7}, { 0, 0, 5, 8}, { 32, 0, 5, 10}, { 0, 0, 5, 11}, { 0, 0, 6, 13}, { 32, 1, 5, 16}, { 0, 1, 5, 18}, { 32, 1, 5, 22}, { 0, 2, 5, 24}, { 32, 3, 5, 32}, { 0, 3, 5, 40}, { 0, 6, 4, 64}, { 16, 6, 4, 64}, { 32, 7, 5, 128}, { 0, 9, 6, 512}, { 0, 11, 6, 2048}, { 48, 0, 4, 0}, { 16, 0, 4, 1}, { 32, 0, 5, 2}, { 32, 0, 5, 3}, { 32, 0, 5, 5}, { 32, 0, 5, 6}, { 32, 0, 5, 8}, { 32, 0, 5, 9}, { 32, 0, 5, 11}, { 32, 0, 5, 12}, { 0, 0, 6, 15}, { 32, 1, 5, 18}, { 32, 1, 5, 20}, { 32, 2, 5, 24}, { 32, 2, 5, 28}, { 32, 3, 5, 40}, { 32, 4, 5, 48}, { 0, 16, 6,65536}, { 0, 15, 6,32768}, { 0, 14, 6,16384}, { 0, 13, 6, 8192}, }; /* LL_defaultDTable */ /* Default FSE distribution table for Offset Codes */ static const ZSTD_seqSymbol OF_defaultDTable[(1<<OF_DEFAULTNORMLOG)+1] = { { 1, 1, 1, OF_DEFAULTNORMLOG}, /* header : fastMode, tableLog */ /* nextState, nbAddBits, nbBits, baseVal */ { 0, 0, 5, 0}, { 0, 6, 4, 61}, { 0, 9, 5, 509}, { 0, 15, 5,32765}, { 0, 21, 5,2097149}, { 0, 3, 5, 5}, { 0, 7, 4, 125}, { 0, 12, 5, 4093}, { 0, 18, 5,262141}, { 0, 23, 5,8388605}, { 0, 5, 5, 29}, { 0, 8, 4, 253}, { 0, 14, 5,16381}, { 0, 20, 5,1048573}, { 0, 2, 5, 1}, { 16, 7, 4, 125}, { 0, 11, 5, 2045}, { 0, 17, 5,131069}, { 0, 22, 5,4194301}, { 0, 4, 5, 13}, { 16, 8, 4, 253}, { 0, 13, 5, 8189}, { 0, 19, 5,524285}, { 0, 1, 5, 1}, { 16, 6, 4, 61}, { 0, 10, 5, 1021}, { 0, 16, 5,65533}, { 0, 28, 5,268435453}, { 0, 27, 5,134217725}, { 0, 26, 5,67108861}, { 0, 25, 5,33554429}, { 0, 24, 5,16777213}, }; /* OF_defaultDTable */ /* Default FSE distribution table for Match Lengths */ static const ZSTD_seqSymbol ML_defaultDTable[(1<<ML_DEFAULTNORMLOG)+1] = { { 1, 1, 1, ML_DEFAULTNORMLOG}, /* header : fastMode, tableLog */ /* nextState, nbAddBits, nbBits, baseVal */ { 0, 0, 6, 3}, { 0, 0, 4, 4}, { 32, 0, 5, 5}, { 0, 0, 5, 6}, { 0, 0, 5, 8}, { 0, 0, 5, 9}, { 0, 0, 5, 11}, { 0, 0, 6, 13}, { 0, 0, 6, 16}, { 0, 0, 6, 19}, { 0, 0, 6, 22}, { 0, 0, 6, 25}, { 0, 0, 6, 28}, { 0, 0, 6, 31}, { 0, 0, 6, 34}, { 0, 1, 6, 37}, { 0, 1, 6, 41}, { 0, 2, 6, 47}, { 0, 3, 6, 59}, { 0, 4, 6, 83}, { 0, 7, 6, 131}, { 0, 9, 6, 515}, { 16, 0, 4, 4}, { 0, 0, 4, 5}, { 32, 0, 5, 6}, { 0, 0, 5, 7}, { 32, 0, 5, 9}, { 0, 0, 5, 10}, { 0, 0, 6, 12}, { 0, 0, 6, 15}, { 0, 0, 6, 18}, { 0, 0, 6, 21}, { 0, 0, 6, 24}, { 0, 0, 6, 27}, { 0, 0, 6, 30}, { 0, 0, 6, 33}, { 0, 1, 6, 35}, { 0, 1, 6, 39}, { 0, 2, 6, 43}, { 0, 3, 6, 51}, { 0, 4, 6, 67}, { 0, 5, 6, 99}, { 0, 8, 6, 259}, { 32, 0, 4, 4}, { 48, 0, 4, 4}, { 16, 0, 4, 5}, { 32, 0, 5, 7}, { 32, 0, 5, 8}, { 32, 0, 5, 10}, { 32, 0, 5, 11}, { 0, 0, 6, 14}, { 0, 0, 6, 17}, { 0, 0, 6, 20}, { 0, 0, 6, 23}, { 0, 0, 6, 26}, { 0, 0, 6, 29}, { 0, 0, 6, 32}, { 0, 16, 6,65539}, { 0, 15, 6,32771}, { 0, 14, 6,16387}, { 0, 13, 6, 8195}, { 0, 12, 6, 4099}, { 0, 11, 6, 2051}, { 0, 10, 6, 1027}, }; /* ML_defaultDTable */ static void ZSTD_buildSeqTable_rle(ZSTD_seqSymbol* dt, U32 baseValue, U8 nbAddBits) { void* ptr = dt; ZSTD_seqSymbol_header* const DTableH = (ZSTD_seqSymbol_header*)ptr; ZSTD_seqSymbol* const cell = dt + 1; DTableH->tableLog = 0; DTableH->fastMode = 0; cell->nbBits = 0; cell->nextState = 0; assert(nbAddBits < 255); cell->nbAdditionalBits = nbAddBits; cell->baseValue = baseValue; } /* ZSTD_buildFSETable() : * generate FSE decoding table for one symbol (ll, ml or off) * cannot fail if input is valid => * all inputs are presumed validated at this stage */ FORCE_INLINE_TEMPLATE void ZSTD_buildFSETable_body(ZSTD_seqSymbol* dt, const short* normalizedCounter, unsigned maxSymbolValue, const U32* baseValue, const U8* nbAdditionalBits, unsigned tableLog, void* wksp, size_t wkspSize) { ZSTD_seqSymbol* const tableDecode = dt+1; U32 const maxSV1 = maxSymbolValue + 1; U32 const tableSize = 1 << tableLog; U16* symbolNext = (U16*)wksp; BYTE* spread = (BYTE*)(symbolNext + MaxSeq + 1); U32 highThreshold = tableSize - 1; /* Sanity Checks */ assert(maxSymbolValue <= MaxSeq); assert(tableLog <= MaxFSELog); assert(wkspSize >= ZSTD_BUILD_FSE_TABLE_WKSP_SIZE); (void)wkspSize; /* Init, lay down lowprob symbols */ { ZSTD_seqSymbol_header DTableH; DTableH.tableLog = tableLog; DTableH.fastMode = 1; { S16 const largeLimit= (S16)(1 << (tableLog-1)); U32 s; for (s=0; s<maxSV1; s++) { if (normalizedCounter[s]==-1) { tableDecode[highThreshold--].baseValue = s; symbolNext[s] = 1; } else { if (normalizedCounter[s] >= largeLimit) DTableH.fastMode=0; assert(normalizedCounter[s]>=0); symbolNext[s] = (U16)normalizedCounter[s]; } } } ZSTD_memcpy(dt, &DTableH, sizeof(DTableH)); } /* Spread symbols */ assert(tableSize <= 512); /* Specialized symbol spreading for the case when there are * no low probability (-1 count) symbols. When compressing * small blocks we avoid low probability symbols to hit this * case, since header decoding speed matters more. */ if (highThreshold == tableSize - 1) { size_t const tableMask = tableSize-1; size_t const step = FSE_TABLESTEP(tableSize); /* First lay down the symbols in order. * We use a uint64_t to lay down 8 bytes at a time. This reduces branch * misses since small blocks generally have small table logs, so nearly * all symbols have counts <= 8. We ensure we have 8 bytes at the end of * our buffer to handle the over-write. */ { U64 const add = 0x0101010101010101ull; size_t pos = 0; U64 sv = 0; U32 s; for (s=0; s<maxSV1; ++s, sv += add) { int i; int const n = normalizedCounter[s]; MEM_write64(spread + pos, sv); for (i = 8; i < n; i += 8) { MEM_write64(spread + pos + i, sv); } pos += n; } } /* Now we spread those positions across the table. * The benefit of doing it in two stages is that we avoid the the * variable size inner loop, which caused lots of branch misses. * Now we can run through all the positions without any branch misses. * We unroll the loop twice, since that is what emperically worked best. */ { size_t position = 0; size_t s; size_t const unroll = 2; assert(tableSize % unroll == 0); /* FSE_MIN_TABLELOG is 5 */ for (s = 0; s < (size_t)tableSize; s += unroll) { size_t u; for (u = 0; u < unroll; ++u) { size_t const uPosition = (position + (u * step)) & tableMask; tableDecode[uPosition].baseValue = spread[s + u]; } position = (position + (unroll * step)) & tableMask; } assert(position == 0); } } else { U32 const tableMask = tableSize-1; U32 const step = FSE_TABLESTEP(tableSize); U32 s, position = 0; for (s=0; s<maxSV1; s++) { int i; int const n = normalizedCounter[s]; for (i=0; i<n; i++) { tableDecode[position].baseValue = s; position = (position + step) & tableMask; while (position > highThreshold) position = (position + step) & tableMask; /* lowprob area */ } } assert(position == 0); /* position must reach all cells once, otherwise normalizedCounter is incorrect */ } /* Build Decoding table */ { U32 u; for (u=0; u<tableSize; u++) { U32 const symbol = tableDecode[u].baseValue; U32 const nextState = symbolNext[symbol]++; tableDecode[u].nbBits = (BYTE) (tableLog - BIT_highbit32(nextState) ); tableDecode[u].nextState = (U16) ( (nextState << tableDecode[u].nbBits) - tableSize); assert(nbAdditionalBits[symbol] < 255); tableDecode[u].nbAdditionalBits = nbAdditionalBits[symbol]; tableDecode[u].baseValue = baseValue[symbol]; } } } /* Avoids the FORCE_INLINE of the _body() function. */ static void ZSTD_buildFSETable_body_default(ZSTD_seqSymbol* dt, const short* normalizedCounter, unsigned maxSymbolValue, const U32* baseValue, const U8* nbAdditionalBits, unsigned tableLog, void* wksp, size_t wkspSize) { ZSTD_buildFSETable_body(dt, normalizedCounter, maxSymbolValue, baseValue, nbAdditionalBits, tableLog, wksp, wkspSize); } #if DYNAMIC_BMI2 BMI2_TARGET_ATTRIBUTE static void ZSTD_buildFSETable_body_bmi2(ZSTD_seqSymbol* dt, const short* normalizedCounter, unsigned maxSymbolValue, const U32* baseValue, const U8* nbAdditionalBits, unsigned tableLog, void* wksp, size_t wkspSize) { ZSTD_buildFSETable_body(dt, normalizedCounter, maxSymbolValue, baseValue, nbAdditionalBits, tableLog, wksp, wkspSize); } #endif void ZSTD_buildFSETable(ZSTD_seqSymbol* dt, const short* normalizedCounter, unsigned maxSymbolValue, const U32* baseValue, const U8* nbAdditionalBits, unsigned tableLog, void* wksp, size_t wkspSize, int bmi2) { #if DYNAMIC_BMI2 if (bmi2) { ZSTD_buildFSETable_body_bmi2(dt, normalizedCounter, maxSymbolValue, baseValue, nbAdditionalBits, tableLog, wksp, wkspSize); return; } #endif (void)bmi2; ZSTD_buildFSETable_body_default(dt, normalizedCounter, maxSymbolValue, baseValue, nbAdditionalBits, tableLog, wksp, wkspSize); } /*! ZSTD_buildSeqTable() : * @return : nb bytes read from src, * or an error code if it fails */ static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymbol** DTablePtr, symbolEncodingType_e type, unsigned max, U32 maxLog, const void* src, size_t srcSize, const U32* baseValue, const U8* nbAdditionalBits, const ZSTD_seqSymbol* defaultTable, U32 flagRepeatTable, int ddictIsCold, int nbSeq, U32* wksp, size_t wkspSize, int bmi2) { switch(type) { case set_rle : RETURN_ERROR_IF(!srcSize, srcSize_wrong, ""); RETURN_ERROR_IF((*(const BYTE*)src) > max, corruption_detected, ""); { U32 const symbol = *(const BYTE*)src; U32 const baseline = baseValue[symbol]; U8 const nbBits = nbAdditionalBits[symbol]; ZSTD_buildSeqTable_rle(DTableSpace, baseline, nbBits); } *DTablePtr = DTableSpace; return 1; case set_basic : *DTablePtr = defaultTable; return 0; case set_repeat: RETURN_ERROR_IF(!flagRepeatTable, corruption_detected, ""); /* prefetch FSE table if used */ if (ddictIsCold && (nbSeq > 24 /* heuristic */)) { const void* const pStart = *DTablePtr; size_t const pSize = sizeof(ZSTD_seqSymbol) * (SEQSYMBOL_TABLE_SIZE(maxLog)); PREFETCH_AREA(pStart, pSize); } return 0; case set_compressed : { unsigned tableLog; S16 norm[MaxSeq+1]; size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize); RETURN_ERROR_IF(FSE_isError(headerSize), corruption_detected, ""); RETURN_ERROR_IF(tableLog > maxLog, corruption_detected, ""); ZSTD_buildFSETable(DTableSpace, norm, max, baseValue, nbAdditionalBits, tableLog, wksp, wkspSize, bmi2); *DTablePtr = DTableSpace; return headerSize; } default : assert(0); RETURN_ERROR(GENERIC, "impossible"); } } size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr, const void* src, size_t srcSize) { const BYTE* const istart = (const BYTE*)src; const BYTE* const iend = istart + srcSize; const BYTE* ip = istart; int nbSeq; DEBUGLOG(5, "ZSTD_decodeSeqHeaders"); /* check */ RETURN_ERROR_IF(srcSize < MIN_SEQUENCES_SIZE, srcSize_wrong, ""); /* SeqHead */ nbSeq = *ip++; if (!nbSeq) { *nbSeqPtr=0; RETURN_ERROR_IF(srcSize != 1, srcSize_wrong, ""); return 1; } if (nbSeq > 0x7F) { if (nbSeq == 0xFF) { RETURN_ERROR_IF(ip+2 > iend, srcSize_wrong, ""); nbSeq = MEM_readLE16(ip) + LONGNBSEQ; ip+=2; } else { RETURN_ERROR_IF(ip >= iend, srcSize_wrong, ""); nbSeq = ((nbSeq-0x80)<<8) + *ip++; } } *nbSeqPtr = nbSeq; /* FSE table descriptors */ RETURN_ERROR_IF(ip+1 > iend, srcSize_wrong, ""); /* minimum possible size: 1 byte for symbol encoding types */ { symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6); symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3); symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3); ip++; /* Build DTables */ { size_t const llhSize = ZSTD_buildSeqTable(dctx->entropy.LLTable, &dctx->LLTptr, LLtype, MaxLL, LLFSELog, ip, iend-ip, LL_base, LL_bits, LL_defaultDTable, dctx->fseEntropy, dctx->ddictIsCold, nbSeq, dctx->workspace, sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx)); RETURN_ERROR_IF(ZSTD_isError(llhSize), corruption_detected, "ZSTD_buildSeqTable failed"); ip += llhSize; } { size_t const ofhSize = ZSTD_buildSeqTable(dctx->entropy.OFTable, &dctx->OFTptr, OFtype, MaxOff, OffFSELog, ip, iend-ip, OF_base, OF_bits, OF_defaultDTable, dctx->fseEntropy, dctx->ddictIsCold, nbSeq, dctx->workspace, sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx)); RETURN_ERROR_IF(ZSTD_isError(ofhSize), corruption_detected, "ZSTD_buildSeqTable failed"); ip += ofhSize; } { size_t const mlhSize = ZSTD_buildSeqTable(dctx->entropy.MLTable, &dctx->MLTptr, MLtype, MaxML, MLFSELog, ip, iend-ip, ML_base, ML_bits, ML_defaultDTable, dctx->fseEntropy, dctx->ddictIsCold, nbSeq, dctx->workspace, sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx)); RETURN_ERROR_IF(ZSTD_isError(mlhSize), corruption_detected, "ZSTD_buildSeqTable failed"); ip += mlhSize; } } return ip-istart; } typedef struct { size_t litLength; size_t matchLength; size_t offset; } seq_t; typedef struct { size_t state; const ZSTD_seqSymbol* table; } ZSTD_fseState; typedef struct { BIT_DStream_t DStream; ZSTD_fseState stateLL; ZSTD_fseState stateOffb; ZSTD_fseState stateML; size_t prevOffset[ZSTD_REP_NUM]; } seqState_t; /*! ZSTD_overlapCopy8() : * Copies 8 bytes from ip to op and updates op and ip where ip <= op. * If the offset is < 8 then the offset is spread to at least 8 bytes. * * Precondition: *ip <= *op * Postcondition: *op - *op >= 8 */ HINT_INLINE void ZSTD_overlapCopy8(BYTE** op, BYTE const** ip, size_t offset) { assert(*ip <= *op); if (offset < 8) { /* close range match, overlap */ static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 }; /* added */ static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 }; /* subtracted */ int const sub2 = dec64table[offset]; (*op)[0] = (*ip)[0]; (*op)[1] = (*ip)[1]; (*op)[2] = (*ip)[2]; (*op)[3] = (*ip)[3]; *ip += dec32table[offset]; ZSTD_copy4(*op+4, *ip); *ip -= sub2; } else { ZSTD_copy8(*op, *ip); } *ip += 8; *op += 8; assert(*op - *ip >= 8); } /*! ZSTD_safecopy() : * Specialized version of memcpy() that is allowed to READ up to WILDCOPY_OVERLENGTH past the input buffer * and write up to 16 bytes past oend_w (op >= oend_w is allowed). * This function is only called in the uncommon case where the sequence is near the end of the block. It * should be fast for a single long sequence, but can be slow for several short sequences. * * @param ovtype controls the overlap detection * - ZSTD_no_overlap: The source and destination are guaranteed to be at least WILDCOPY_VECLEN bytes apart. * - ZSTD_overlap_src_before_dst: The src and dst may overlap and may be any distance apart. * The src buffer must be before the dst buffer. */ static void ZSTD_safecopy(BYTE* op, const BYTE* const oend_w, BYTE const* ip, ptrdiff_t length, ZSTD_overlap_e ovtype) { ptrdiff_t const diff = op - ip; BYTE* const oend = op + length; assert((ovtype == ZSTD_no_overlap && (diff <= -8 || diff >= 8 || op >= oend_w)) || (ovtype == ZSTD_overlap_src_before_dst && diff >= 0)); if (length < 8) { /* Handle short lengths. */ while (op < oend) *op++ = *ip++; return; } if (ovtype == ZSTD_overlap_src_before_dst) { /* Copy 8 bytes and ensure the offset >= 8 when there can be overlap. */ assert(length >= 8); ZSTD_overlapCopy8(&op, &ip, diff); length -= 8; assert(op - ip >= 8); assert(op <= oend); } if (oend <= oend_w) { /* No risk of overwrite. */ ZSTD_wildcopy(op, ip, length, ovtype); return; } if (op <= oend_w) { /* Wildcopy until we get close to the end. */ assert(oend > oend_w); ZSTD_wildcopy(op, ip, oend_w - op, ovtype); ip += oend_w - op; op += oend_w - op; } /* Handle the leftovers. */ while (op < oend) *op++ = *ip++; } /* ZSTD_safecopyDstBeforeSrc(): * This version allows overlap with dst before src, or handles the non-overlap case with dst after src * Kept separate from more common ZSTD_safecopy case to avoid performance impact to the safecopy common case */ static void ZSTD_safecopyDstBeforeSrc(BYTE* op, BYTE const* ip, ptrdiff_t length) { ptrdiff_t const diff = op - ip; BYTE* const oend = op + length; if (length < 8 || diff > -8) { /* Handle short lengths, close overlaps, and dst not before src. */ while (op < oend) *op++ = *ip++; return; } if (op <= oend - WILDCOPY_OVERLENGTH && diff < -WILDCOPY_VECLEN) { ZSTD_wildcopy(op, ip, oend - WILDCOPY_OVERLENGTH - op, ZSTD_no_overlap); ip += oend - WILDCOPY_OVERLENGTH - op; op += oend - WILDCOPY_OVERLENGTH - op; } /* Handle the leftovers. */ while (op < oend) *op++ = *ip++; } /* ZSTD_execSequenceEnd(): * This version handles cases that are near the end of the output buffer. It requires * more careful checks to make sure there is no overflow. By separating out these hard * and unlikely cases, we can speed up the common cases. * * NOTE: This function needs to be fast for a single long sequence, but doesn't need * to be optimized for many small sequences, since those fall into ZSTD_execSequence(). */ FORCE_NOINLINE size_t ZSTD_execSequenceEnd(BYTE* op, BYTE* const oend, seq_t sequence, const BYTE** litPtr, const BYTE* const litLimit, const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd) { BYTE* const oLitEnd = op + sequence.litLength; size_t const sequenceLength = sequence.litLength + sequence.matchLength; const BYTE* const iLitEnd = *litPtr + sequence.litLength; const BYTE* match = oLitEnd - sequence.offset; BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH; /* bounds checks : careful of address space overflow in 32-bit mode */ RETURN_ERROR_IF(sequenceLength > (size_t)(oend - op), dstSize_tooSmall, "last match must fit within dstBuffer"); RETURN_ERROR_IF(sequence.litLength > (size_t)(litLimit - *litPtr), corruption_detected, "try to read beyond literal buffer"); assert(op < op + sequenceLength); assert(oLitEnd < op + sequenceLength); /* copy literals */ ZSTD_safecopy(op, oend_w, *litPtr, sequence.litLength, ZSTD_no_overlap); op = oLitEnd; *litPtr = iLitEnd; /* copy Match */ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) { /* offset beyond prefix */ RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected, ""); match = dictEnd - (prefixStart - match); if (match + sequence.matchLength <= dictEnd) { ZSTD_memmove(oLitEnd, match, sequence.matchLength); return sequenceLength; } /* span extDict & currentPrefixSegment */ { size_t const length1 = dictEnd - match; ZSTD_memmove(oLitEnd, match, length1); op = oLitEnd + length1; sequence.matchLength -= length1; match = prefixStart; } } ZSTD_safecopy(op, oend_w, match, sequence.matchLength, ZSTD_overlap_src_before_dst); return sequenceLength; } /* ZSTD_execSequenceEndSplitLitBuffer(): * This version is intended to be used during instances where the litBuffer is still split. It is kept separate to avoid performance impact for the good case. */ FORCE_NOINLINE size_t ZSTD_execSequenceEndSplitLitBuffer(BYTE* op, BYTE* const oend, const BYTE* const oend_w, seq_t sequence, const BYTE** litPtr, const BYTE* const litLimit, const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd) { BYTE* const oLitEnd = op + sequence.litLength; size_t const sequenceLength = sequence.litLength + sequence.matchLength; const BYTE* const iLitEnd = *litPtr + sequence.litLength; const BYTE* match = oLitEnd - sequence.offset; /* bounds checks : careful of address space overflow in 32-bit mode */ RETURN_ERROR_IF(sequenceLength > (size_t)(oend - op), dstSize_tooSmall, "last match must fit within dstBuffer"); RETURN_ERROR_IF(sequence.litLength > (size_t)(litLimit - *litPtr), corruption_detected, "try to read beyond literal buffer"); assert(op < op + sequenceLength); assert(oLitEnd < op + sequenceLength); /* copy literals */ RETURN_ERROR_IF(op > *litPtr && op < *litPtr + sequence.litLength, dstSize_tooSmall, "output should not catch up to and overwrite literal buffer"); ZSTD_safecopyDstBeforeSrc(op, *litPtr, sequence.litLength); op = oLitEnd; *litPtr = iLitEnd; /* copy Match */ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) { /* offset beyond prefix */ RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected, ""); match = dictEnd - (prefixStart - match); if (match + sequence.matchLength <= dictEnd) { ZSTD_memmove(oLitEnd, match, sequence.matchLength); return sequenceLength; } /* span extDict & currentPrefixSegment */ { size_t const length1 = dictEnd - match; ZSTD_memmove(oLitEnd, match, length1); op = oLitEnd + length1; sequence.matchLength -= length1; match = prefixStart; } } ZSTD_safecopy(op, oend_w, match, sequence.matchLength, ZSTD_overlap_src_before_dst); return sequenceLength; } HINT_INLINE size_t ZSTD_execSequence(BYTE* op, BYTE* const oend, seq_t sequence, const BYTE** litPtr, const BYTE* const litLimit, const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd) { BYTE* const oLitEnd = op + sequence.litLength; size_t const sequenceLength = sequence.litLength + sequence.matchLength; BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */ BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH; /* risk : address space underflow on oend=NULL */ const BYTE* const iLitEnd = *litPtr + sequence.litLength; const BYTE* match = oLitEnd - sequence.offset; assert(op != NULL /* Precondition */); assert(oend_w < oend /* No underflow */); /* Handle edge cases in a slow path: * - Read beyond end of literals * - Match end is within WILDCOPY_OVERLIMIT of oend * - 32-bit mode and the match length overflows */ if (UNLIKELY( iLitEnd > litLimit || oMatchEnd > oend_w || (MEM_32bits() && (size_t)(oend - op) < sequenceLength + WILDCOPY_OVERLENGTH))) return ZSTD_execSequenceEnd(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd); /* Assumptions (everything else goes into ZSTD_execSequenceEnd()) */ assert(op <= oLitEnd /* No overflow */); assert(oLitEnd < oMatchEnd /* Non-zero match & no overflow */); assert(oMatchEnd <= oend /* No underflow */); assert(iLitEnd <= litLimit /* Literal length is in bounds */); assert(oLitEnd <= oend_w /* Can wildcopy literals */); assert(oMatchEnd <= oend_w /* Can wildcopy matches */); /* Copy Literals: * Split out litLength <= 16 since it is nearly always true. +1.6% on gcc-9. * We likely don't need the full 32-byte wildcopy. */ assert(WILDCOPY_OVERLENGTH >= 16); ZSTD_copy16(op, (*litPtr)); if (UNLIKELY(sequence.litLength > 16)) { ZSTD_wildcopy(op + 16, (*litPtr) + 16, sequence.litLength - 16, ZSTD_no_overlap); } op = oLitEnd; *litPtr = iLitEnd; /* update for next sequence */ /* Copy Match */ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) { /* offset beyond prefix -> go into extDict */ RETURN_ERROR_IF(UNLIKELY(sequence.offset > (size_t)(oLitEnd - virtualStart)), corruption_detected, ""); match = dictEnd + (match - prefixStart); if (match + sequence.matchLength <= dictEnd) { ZSTD_memmove(oLitEnd, match, sequence.matchLength); return sequenceLength; } /* span extDict & currentPrefixSegment */ { size_t const length1 = dictEnd - match; ZSTD_memmove(oLitEnd, match, length1); op = oLitEnd + length1; sequence.matchLength -= length1; match = prefixStart; } } /* Match within prefix of 1 or more bytes */ assert(op <= oMatchEnd); assert(oMatchEnd <= oend_w); assert(match >= prefixStart); assert(sequence.matchLength >= 1); /* Nearly all offsets are >= WILDCOPY_VECLEN bytes, which means we can use wildcopy * without overlap checking. */ if (LIKELY(sequence.offset >= WILDCOPY_VECLEN)) { /* We bet on a full wildcopy for matches, since we expect matches to be * longer than literals (in general). In silesia, ~10% of matches are longer * than 16 bytes. */ ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength, ZSTD_no_overlap); return sequenceLength; } assert(sequence.offset < WILDCOPY_VECLEN); /* Copy 8 bytes and spread the offset to be >= 8. */ ZSTD_overlapCopy8(&op, &match, sequence.offset); /* If the match length is > 8 bytes, then continue with the wildcopy. */ if (sequence.matchLength > 8) { assert(op < oMatchEnd); ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength - 8, ZSTD_overlap_src_before_dst); } return sequenceLength; } HINT_INLINE size_t ZSTD_execSequenceSplitLitBuffer(BYTE* op, BYTE* const oend, const BYTE* const oend_w, seq_t sequence, const BYTE** litPtr, const BYTE* const litLimit, const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd) { BYTE* const oLitEnd = op + sequence.litLength; size_t const sequenceLength = sequence.litLength + sequence.matchLength; BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */ const BYTE* const iLitEnd = *litPtr + sequence.litLength; const BYTE* match = oLitEnd - sequence.offset; assert(op != NULL /* Precondition */); assert(oend_w < oend /* No underflow */); /* Handle edge cases in a slow path: * - Read beyond end of literals * - Match end is within WILDCOPY_OVERLIMIT of oend * - 32-bit mode and the match length overflows */ if (UNLIKELY( iLitEnd > litLimit || oMatchEnd > oend_w || (MEM_32bits() && (size_t)(oend - op) < sequenceLength + WILDCOPY_OVERLENGTH))) return ZSTD_execSequenceEndSplitLitBuffer(op, oend, oend_w, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd); /* Assumptions (everything else goes into ZSTD_execSequenceEnd()) */ assert(op <= oLitEnd /* No overflow */); assert(oLitEnd < oMatchEnd /* Non-zero match & no overflow */); assert(oMatchEnd <= oend /* No underflow */); assert(iLitEnd <= litLimit /* Literal length is in bounds */); assert(oLitEnd <= oend_w /* Can wildcopy literals */); assert(oMatchEnd <= oend_w /* Can wildcopy matches */); /* Copy Literals: * Split out litLength <= 16 since it is nearly always true. +1.6% on gcc-9. * We likely don't need the full 32-byte wildcopy. */ assert(WILDCOPY_OVERLENGTH >= 16); ZSTD_copy16(op, (*litPtr)); if (UNLIKELY(sequence.litLength > 16)) { ZSTD_wildcopy(op+16, (*litPtr)+16, sequence.litLength-16, ZSTD_no_overlap); } op = oLitEnd; *litPtr = iLitEnd; /* update for next sequence */ /* Copy Match */ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) { /* offset beyond prefix -> go into extDict */ RETURN_ERROR_IF(UNLIKELY(sequence.offset > (size_t)(oLitEnd - virtualStart)), corruption_detected, ""); match = dictEnd + (match - prefixStart); if (match + sequence.matchLength <= dictEnd) { ZSTD_memmove(oLitEnd, match, sequence.matchLength); return sequenceLength; } /* span extDict & currentPrefixSegment */ { size_t const length1 = dictEnd - match; ZSTD_memmove(oLitEnd, match, length1); op = oLitEnd + length1; sequence.matchLength -= length1; match = prefixStart; } } /* Match within prefix of 1 or more bytes */ assert(op <= oMatchEnd); assert(oMatchEnd <= oend_w); assert(match >= prefixStart); assert(sequence.matchLength >= 1); /* Nearly all offsets are >= WILDCOPY_VECLEN bytes, which means we can use wildcopy * without overlap checking. */ if (LIKELY(sequence.offset >= WILDCOPY_VECLEN)) { /* We bet on a full wildcopy for matches, since we expect matches to be * longer than literals (in general). In silesia, ~10% of matches are longer * than 16 bytes. */ ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength, ZSTD_no_overlap); return sequenceLength; } assert(sequence.offset < WILDCOPY_VECLEN); /* Copy 8 bytes and spread the offset to be >= 8. */ ZSTD_overlapCopy8(&op, &match, sequence.offset); /* If the match length is > 8 bytes, then continue with the wildcopy. */ if (sequence.matchLength > 8) { assert(op < oMatchEnd); ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8, ZSTD_overlap_src_before_dst); } return sequenceLength; } static void ZSTD_initFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, const ZSTD_seqSymbol* dt) { const void* ptr = dt; const ZSTD_seqSymbol_header* const DTableH = (const ZSTD_seqSymbol_header*)ptr; DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog); DEBUGLOG(6, "ZSTD_initFseState : val=%u using %u bits", (U32)DStatePtr->state, DTableH->tableLog); BIT_reloadDStream(bitD); DStatePtr->table = dt + 1; } FORCE_INLINE_TEMPLATE void ZSTD_updateFseStateWithDInfo(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, U16 nextState, U32 nbBits) { size_t const lowBits = BIT_readBits(bitD, nbBits); DStatePtr->state = nextState + lowBits; } /* We need to add at most (ZSTD_WINDOWLOG_MAX_32 - 1) bits to read the maximum * offset bits. But we can only read at most (STREAM_ACCUMULATOR_MIN_32 - 1) * bits before reloading. This value is the maximum number of bytes we read * after reloading when we are decoding long offsets. */ #define LONG_OFFSETS_MAX_EXTRA_BITS_32 \ (ZSTD_WINDOWLOG_MAX_32 > STREAM_ACCUMULATOR_MIN_32 \ ? ZSTD_WINDOWLOG_MAX_32 - STREAM_ACCUMULATOR_MIN_32 \ : 0) typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e; FORCE_INLINE_TEMPLATE seq_t ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets) { seq_t seq; const ZSTD_seqSymbol* const llDInfo = seqState->stateLL.table + seqState->stateLL.state; const ZSTD_seqSymbol* const mlDInfo = seqState->stateML.table + seqState->stateML.state; const ZSTD_seqSymbol* const ofDInfo = seqState->stateOffb.table + seqState->stateOffb.state; seq.matchLength = mlDInfo->baseValue; seq.litLength = llDInfo->baseValue; { U32 const ofBase = ofDInfo->baseValue; BYTE const llBits = llDInfo->nbAdditionalBits; BYTE const mlBits = mlDInfo->nbAdditionalBits; BYTE const ofBits = ofDInfo->nbAdditionalBits; BYTE const totalBits = llBits+mlBits+ofBits; U16 const llNext = llDInfo->nextState; U16 const mlNext = mlDInfo->nextState; U16 const ofNext = ofDInfo->nextState; U32 const llnbBits = llDInfo->nbBits; U32 const mlnbBits = mlDInfo->nbBits; U32 const ofnbBits = ofDInfo->nbBits; /* * As gcc has better branch and block analyzers, sometimes it is only * valuable to mark likelyness for clang, it gives around 3-4% of * performance. */ /* sequence */ { size_t offset; #if defined(__clang__) if (LIKELY(ofBits > 1)) { #else if (ofBits > 1) { #endif ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1); ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5); assert(ofBits <= MaxOff); if (MEM_32bits() && longOffsets && (ofBits >= STREAM_ACCUMULATOR_MIN_32)) { U32 const extraBits = ofBits - MIN(ofBits, 32 - seqState->DStream.bitsConsumed); offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits); BIT_reloadDStream(&seqState->DStream); if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits); assert(extraBits <= LONG_OFFSETS_MAX_EXTRA_BITS_32); /* to avoid another reload */ } else { offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits/*>0*/); /* <= (ZSTD_WINDOWLOG_MAX-1) bits */ if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); } seqState->prevOffset[2] = seqState->prevOffset[1]; seqState->prevOffset[1] = seqState->prevOffset[0]; seqState->prevOffset[0] = offset; } else { U32 const ll0 = (llDInfo->baseValue == 0); if (LIKELY((ofBits == 0))) { offset = seqState->prevOffset[ll0]; seqState->prevOffset[1] = seqState->prevOffset[!ll0]; seqState->prevOffset[0] = offset; } else { offset = ofBase + ll0 + BIT_readBitsFast(&seqState->DStream, 1); { size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset]; temp += !temp; /* 0 is not valid; input is corrupted; force offset to 1 */ if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1]; seqState->prevOffset[1] = seqState->prevOffset[0]; seqState->prevOffset[0] = offset = temp; } } } seq.offset = offset; } #if defined(__clang__) if (UNLIKELY(mlBits > 0)) #else if (mlBits > 0) #endif seq.matchLength += BIT_readBitsFast(&seqState->DStream, mlBits/*>0*/); if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32)) BIT_reloadDStream(&seqState->DStream); if (MEM_64bits() && UNLIKELY(totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog))) BIT_reloadDStream(&seqState->DStream); /* Ensure there are enough bits to read the rest of data in 64-bit mode. */ ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64); #if defined(__clang__) if (UNLIKELY(llBits > 0)) #else if (llBits > 0) #endif seq.litLength += BIT_readBitsFast(&seqState->DStream, llBits/*>0*/); if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u", (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset); ZSTD_updateFseStateWithDInfo(&seqState->stateLL, &seqState->DStream, llNext, llnbBits); /* <= 9 bits */ ZSTD_updateFseStateWithDInfo(&seqState->stateML, &seqState->DStream, mlNext, mlnbBits); /* <= 9 bits */ if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); /* <= 18 bits */ ZSTD_updateFseStateWithDInfo(&seqState->stateOffb, &seqState->DStream, ofNext, ofnbBits); /* <= 8 bits */ } return seq; } #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION MEM_STATIC int ZSTD_dictionaryIsActive(ZSTD_DCtx const* dctx, BYTE const* prefixStart, BYTE const* oLitEnd) { size_t const windowSize = dctx->fParams.windowSize; /* No dictionary used. */ if (dctx->dictContentEndForFuzzing == NULL) return 0; /* Dictionary is our prefix. */ if (prefixStart == dctx->dictContentBeginForFuzzing) return 1; /* Dictionary is not our ext-dict. */ if (dctx->dictEnd != dctx->dictContentEndForFuzzing) return 0; /* Dictionary is not within our window size. */ if ((size_t)(oLitEnd - prefixStart) >= windowSize) return 0; /* Dictionary is active. */ return 1; } MEM_STATIC void ZSTD_assertValidSequence( ZSTD_DCtx const* dctx, BYTE const* op, BYTE const* oend, seq_t const seq, BYTE const* prefixStart, BYTE const* virtualStart) { #if DEBUGLEVEL >= 1 size_t const windowSize = dctx->fParams.windowSize; size_t const sequenceSize = seq.litLength + seq.matchLength; BYTE const* const oLitEnd = op + seq.litLength; DEBUGLOG(6, "Checking sequence: litL=%u matchL=%u offset=%u", (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset); assert(op <= oend); assert((size_t)(oend - op) >= sequenceSize); assert(sequenceSize <= ZSTD_BLOCKSIZE_MAX); if (ZSTD_dictionaryIsActive(dctx, prefixStart, oLitEnd)) { size_t const dictSize = (size_t)((char const*)dctx->dictContentEndForFuzzing - (char const*)dctx->dictContentBeginForFuzzing); /* Offset must be within the dictionary. */ assert(seq.offset <= (size_t)(oLitEnd - virtualStart)); assert(seq.offset <= windowSize + dictSize); } else { /* Offset must be within our window. */ assert(seq.offset <= windowSize); } #else (void)dctx, (void)op, (void)oend, (void)seq, (void)prefixStart, (void)virtualStart; #endif } #endif #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG FORCE_INLINE_TEMPLATE size_t DONT_VECTORIZE ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { const BYTE* ip = (const BYTE*)seqStart; const BYTE* const iend = ip + seqSize; BYTE* const ostart = (BYTE*)dst; BYTE* const oend = ostart + maxDstSize; BYTE* op = ostart; const BYTE* litPtr = dctx->litPtr; const BYTE* litBufferEnd = dctx->litBufferEnd; const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart); const BYTE* const vBase = (const BYTE*) (dctx->virtualStart); const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd); DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer"); (void)frame; /* Regen sequences */ if (nbSeq) { seqState_t seqState; dctx->fseEntropy = 1; { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; } RETURN_ERROR_IF( ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend-ip)), corruption_detected, ""); ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr); ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr); ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr); assert(dst != NULL); ZSTD_STATIC_ASSERT( BIT_DStream_unfinished < BIT_DStream_completed && BIT_DStream_endOfBuffer < BIT_DStream_completed && BIT_DStream_completed < BIT_DStream_overflow); /* decompress without overrunning litPtr begins */ { seq_t sequence = ZSTD_decodeSequence(&seqState, isLongOffset); /* Align the decompression loop to 32 + 16 bytes. * * zstd compiled with gcc-9 on an Intel i9-9900k shows 10% decompression * speed swings based on the alignment of the decompression loop. This * performance swing is caused by parts of the decompression loop falling * out of the DSB. The entire decompression loop should fit in the DSB, * when it can't we get much worse performance. You can measure if you've * hit the good case or the bad case with this perf command for some * compressed file test.zst: * * perf stat -e cycles -e instructions -e idq.all_dsb_cycles_any_uops \ * -e idq.all_mite_cycles_any_uops -- ./zstd -tq test.zst * * If you see most cycles served out of the MITE you've hit the bad case. * If you see most cycles served out of the DSB you've hit the good case. * If it is pretty even then you may be in an okay case. * * This issue has been reproduced on the following CPUs: * - Kabylake: Macbook Pro (15-inch, 2019) 2.4 GHz Intel Core i9 * Use Instruments->Counters to get DSB/MITE cycles. * I never got performance swings, but I was able to * go from the good case of mostly DSB to half of the * cycles served from MITE. * - Coffeelake: Intel i9-9900k * - Coffeelake: Intel i7-9700k * * I haven't been able to reproduce the instability or DSB misses on any * of the following CPUS: * - Haswell * - Broadwell: Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GH * - Skylake * * Alignment is done for each of the three major decompression loops: * - ZSTD_decompressSequences_bodySplitLitBuffer - presplit section of the literal buffer * - ZSTD_decompressSequences_bodySplitLitBuffer - postsplit section of the literal buffer * - ZSTD_decompressSequences_body * Alignment choices are made to minimize large swings on bad cases and influence on performance * from changes external to this code, rather than to overoptimize on the current commit. * * If you are seeing performance stability this script can help test. * It tests on 4 commits in zstd where I saw performance change. * * https://gist.github.com/terrelln/9889fc06a423fd5ca6e99351564473f4 */ #if defined(__x86_64__) __asm__(".p2align 6"); # if __GNUC__ >= 7 /* good for gcc-7, gcc-9, and gcc-11 */ __asm__("nop"); __asm__(".p2align 5"); __asm__("nop"); __asm__(".p2align 4"); # if __GNUC__ == 8 || __GNUC__ == 10 /* good for gcc-8 and gcc-10 */ __asm__("nop"); __asm__(".p2align 3"); # endif # endif #endif /* Handle the initial state where litBuffer is currently split between dst and litExtraBuffer */ for (; litPtr + sequence.litLength <= dctx->litBufferEnd; ) { size_t const oneSeqSize = ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequence.litLength - WILDCOPY_OVERLENGTH, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase); #endif if (UNLIKELY(ZSTD_isError(oneSeqSize))) return oneSeqSize; DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize); op += oneSeqSize; if (UNLIKELY(!--nbSeq)) break; BIT_reloadDStream(&(seqState.DStream)); sequence = ZSTD_decodeSequence(&seqState, isLongOffset); } /* If there are more sequences, they will need to read literals from litExtraBuffer; copy over the remainder from dst and update litPtr and litEnd */ if (nbSeq > 0) { const size_t leftoverLit = dctx->litBufferEnd - litPtr; if (leftoverLit) { RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer"); ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit); sequence.litLength -= leftoverLit; op += leftoverLit; } litPtr = dctx->litExtraBuffer; litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE; dctx->litBufferLocation = ZSTD_not_in_dst; { size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase); #endif if (UNLIKELY(ZSTD_isError(oneSeqSize))) return oneSeqSize; DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize); op += oneSeqSize; if (--nbSeq) BIT_reloadDStream(&(seqState.DStream)); } } } if (nbSeq > 0) /* there is remaining lit from extra buffer */ { #if defined(__x86_64__) __asm__(".p2align 6"); __asm__("nop"); # if __GNUC__ != 7 /* worse for gcc-7 better for gcc-8, gcc-9, and gcc-10 and clang */ __asm__(".p2align 4"); __asm__("nop"); __asm__(".p2align 3"); # elif __GNUC__ >= 11 __asm__(".p2align 3"); # else __asm__(".p2align 5"); __asm__("nop"); __asm__(".p2align 3"); # endif #endif for (; ; ) { seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset); size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase); #endif if (UNLIKELY(ZSTD_isError(oneSeqSize))) return oneSeqSize; DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize); op += oneSeqSize; if (UNLIKELY(!--nbSeq)) break; BIT_reloadDStream(&(seqState.DStream)); } } /* check if reached exact end */ DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer: after decode loop, remaining nbSeq : %i", nbSeq); RETURN_ERROR_IF(nbSeq, corruption_detected, ""); RETURN_ERROR_IF(BIT_reloadDStream(&seqState.DStream) < BIT_DStream_completed, corruption_detected, ""); /* save reps for next block */ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); } } /* last literal segment */ if (dctx->litBufferLocation == ZSTD_split) /* split hasn't been reached yet, first get dst then copy litExtraBuffer */ { size_t const lastLLSize = litBufferEnd - litPtr; RETURN_ERROR_IF(lastLLSize > (size_t)(oend - op), dstSize_tooSmall, ""); if (op != NULL) { ZSTD_memmove(op, litPtr, lastLLSize); op += lastLLSize; } litPtr = dctx->litExtraBuffer; litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE; dctx->litBufferLocation = ZSTD_not_in_dst; } { size_t const lastLLSize = litBufferEnd - litPtr; RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, ""); if (op != NULL) { ZSTD_memcpy(op, litPtr, lastLLSize); op += lastLLSize; } } return op-ostart; } FORCE_INLINE_TEMPLATE size_t DONT_VECTORIZE ZSTD_decompressSequences_body(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { const BYTE* ip = (const BYTE*)seqStart; const BYTE* const iend = ip + seqSize; BYTE* const ostart = (BYTE*)dst; BYTE* const oend = dctx->litBufferLocation == ZSTD_not_in_dst ? ostart + maxDstSize : dctx->litBuffer; BYTE* op = ostart; const BYTE* litPtr = dctx->litPtr; const BYTE* const litEnd = litPtr + dctx->litSize; const BYTE* const prefixStart = (const BYTE*)(dctx->prefixStart); const BYTE* const vBase = (const BYTE*)(dctx->virtualStart); const BYTE* const dictEnd = (const BYTE*)(dctx->dictEnd); DEBUGLOG(5, "ZSTD_decompressSequences_body"); (void)frame; /* Regen sequences */ if (nbSeq) { seqState_t seqState; dctx->fseEntropy = 1; { U32 i; for (i = 0; i < ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; } RETURN_ERROR_IF( ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend - ip)), corruption_detected, ""); ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr); ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr); ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr); assert(dst != NULL); ZSTD_STATIC_ASSERT( BIT_DStream_unfinished < BIT_DStream_completed && BIT_DStream_endOfBuffer < BIT_DStream_completed && BIT_DStream_completed < BIT_DStream_overflow); #if defined(__x86_64__) __asm__(".p2align 6"); __asm__("nop"); # if __GNUC__ >= 7 __asm__(".p2align 5"); __asm__("nop"); __asm__(".p2align 3"); # else __asm__(".p2align 4"); __asm__("nop"); __asm__(".p2align 3"); # endif #endif for ( ; ; ) { seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset); size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, prefixStart, vBase, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase); #endif if (UNLIKELY(ZSTD_isError(oneSeqSize))) return oneSeqSize; DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize); op += oneSeqSize; if (UNLIKELY(!--nbSeq)) break; BIT_reloadDStream(&(seqState.DStream)); } /* check if reached exact end */ DEBUGLOG(5, "ZSTD_decompressSequences_body: after decode loop, remaining nbSeq : %i", nbSeq); RETURN_ERROR_IF(nbSeq, corruption_detected, ""); RETURN_ERROR_IF(BIT_reloadDStream(&seqState.DStream) < BIT_DStream_completed, corruption_detected, ""); /* save reps for next block */ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); } } /* last literal segment */ { size_t const lastLLSize = litEnd - litPtr; RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, ""); if (op != NULL) { ZSTD_memcpy(op, litPtr, lastLLSize); op += lastLLSize; } } return op-ostart; } static size_t ZSTD_decompressSequences_default(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } static size_t ZSTD_decompressSequencesSplitLitBuffer_default(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */ #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT FORCE_INLINE_TEMPLATE size_t ZSTD_prefetchMatch(size_t prefetchPos, seq_t const sequence, const BYTE* const prefixStart, const BYTE* const dictEnd) { prefetchPos += sequence.litLength; { const BYTE* const matchBase = (sequence.offset > prefetchPos) ? dictEnd : prefixStart; const BYTE* const match = matchBase + prefetchPos - sequence.offset; /* note : this operation can overflow when seq.offset is really too large, which can only happen when input is corrupted. * No consequence though : memory address is only used for prefetching, not for dereferencing */ PREFETCH_L1(match); PREFETCH_L1(match+CACHELINE_SIZE); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */ } return prefetchPos + sequence.matchLength; } /* This decoding function employs prefetching * to reduce latency impact of cache misses. * It's generally employed when block contains a significant portion of long-distance matches * or when coupled with a "cold" dictionary */ FORCE_INLINE_TEMPLATE size_t ZSTD_decompressSequencesLong_body( ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { const BYTE* ip = (const BYTE*)seqStart; const BYTE* const iend = ip + seqSize; BYTE* const ostart = (BYTE*)dst; BYTE* const oend = dctx->litBufferLocation == ZSTD_in_dst ? dctx->litBuffer : ostart + maxDstSize; BYTE* op = ostart; const BYTE* litPtr = dctx->litPtr; const BYTE* litBufferEnd = dctx->litBufferEnd; const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart); const BYTE* const dictStart = (const BYTE*) (dctx->virtualStart); const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd); (void)frame; /* Regen sequences */ if (nbSeq) { #define STORED_SEQS 8 #define STORED_SEQS_MASK (STORED_SEQS-1) #define ADVANCED_SEQS STORED_SEQS seq_t sequences[STORED_SEQS]; int const seqAdvance = MIN(nbSeq, ADVANCED_SEQS); seqState_t seqState; int seqNb; size_t prefetchPos = (size_t)(op-prefixStart); /* track position relative to prefixStart */ dctx->fseEntropy = 1; { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; } assert(dst != NULL); assert(iend >= ip); RETURN_ERROR_IF( ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend-ip)), corruption_detected, ""); ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr); ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr); ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr); /* prepare in advance */ for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && (seqNb<seqAdvance); seqNb++) { seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset); prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd); sequences[seqNb] = sequence; } RETURN_ERROR_IF(seqNb<seqAdvance, corruption_detected, ""); /* decompress without stomping litBuffer */ for (; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && (seqNb < nbSeq); seqNb++) { seq_t sequence = ZSTD_decodeSequence(&seqState, isLongOffset); size_t oneSeqSize; if (dctx->litBufferLocation == ZSTD_split && litPtr + sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength > dctx->litBufferEnd) { /* lit buffer is reaching split point, empty out the first buffer and transition to litExtraBuffer */ const size_t leftoverLit = dctx->litBufferEnd - litPtr; if (leftoverLit) { RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer"); ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit); sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength -= leftoverLit; op += leftoverLit; } litPtr = dctx->litExtraBuffer; litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE; dctx->litBufferLocation = ZSTD_not_in_dst; oneSeqSize = ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart); #endif if (ZSTD_isError(oneSeqSize)) return oneSeqSize; prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd); sequences[seqNb & STORED_SEQS_MASK] = sequence; op += oneSeqSize; } else { /* lit buffer is either wholly contained in first or second split, or not split at all*/ oneSeqSize = dctx->litBufferLocation == ZSTD_split ? ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength - WILDCOPY_OVERLENGTH, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd) : ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart); #endif if (ZSTD_isError(oneSeqSize)) return oneSeqSize; prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd); sequences[seqNb & STORED_SEQS_MASK] = sequence; op += oneSeqSize; } } RETURN_ERROR_IF(seqNb<nbSeq, corruption_detected, ""); /* finish queue */ seqNb -= seqAdvance; for ( ; seqNb<nbSeq ; seqNb++) { seq_t *sequence = &(sequences[seqNb&STORED_SEQS_MASK]); if (dctx->litBufferLocation == ZSTD_split && litPtr + sequence->litLength > dctx->litBufferEnd) { const size_t leftoverLit = dctx->litBufferEnd - litPtr; if (leftoverLit) { RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer"); ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit); sequence->litLength -= leftoverLit; op += leftoverLit; } litPtr = dctx->litExtraBuffer; litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE; dctx->litBufferLocation = ZSTD_not_in_dst; { size_t const oneSeqSize = ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart); #endif if (ZSTD_isError(oneSeqSize)) return oneSeqSize; op += oneSeqSize; } } else { size_t const oneSeqSize = dctx->litBufferLocation == ZSTD_split ? ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequence->litLength - WILDCOPY_OVERLENGTH, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd) : ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd); #if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE) assert(!ZSTD_isError(oneSeqSize)); if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart); #endif if (ZSTD_isError(oneSeqSize)) return oneSeqSize; op += oneSeqSize; } } /* save reps for next block */ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); } } /* last literal segment */ if (dctx->litBufferLocation == ZSTD_split) /* first deplete literal buffer in dst, then copy litExtraBuffer */ { size_t const lastLLSize = litBufferEnd - litPtr; RETURN_ERROR_IF(lastLLSize > (size_t)(oend - op), dstSize_tooSmall, ""); if (op != NULL) { ZSTD_memmove(op, litPtr, lastLLSize); op += lastLLSize; } litPtr = dctx->litExtraBuffer; litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE; } { size_t const lastLLSize = litBufferEnd - litPtr; RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, ""); if (op != NULL) { ZSTD_memmove(op, litPtr, lastLLSize); op += lastLLSize; } } return op-ostart; } static size_t ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */ #if DYNAMIC_BMI2 #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG static BMI2_TARGET_ATTRIBUTE size_t DONT_VECTORIZE ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } static BMI2_TARGET_ATTRIBUTE size_t DONT_VECTORIZE ZSTD_decompressSequencesSplitLitBuffer_bmi2(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */ #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT static BMI2_TARGET_ATTRIBUTE size_t ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */ #endif /* DYNAMIC_BMI2 */ typedef size_t (*ZSTD_decompressSequences_t)( ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame); #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG static size_t ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { DEBUGLOG(5, "ZSTD_decompressSequences"); #if DYNAMIC_BMI2 if (ZSTD_DCtx_get_bmi2(dctx)) { return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } static size_t ZSTD_decompressSequencesSplitLitBuffer(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { DEBUGLOG(5, "ZSTD_decompressSequencesSplitLitBuffer"); #if DYNAMIC_BMI2 if (ZSTD_DCtx_get_bmi2(dctx)) { return ZSTD_decompressSequencesSplitLitBuffer_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif return ZSTD_decompressSequencesSplitLitBuffer_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */ #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT /* ZSTD_decompressSequencesLong() : * decompression function triggered when a minimum share of offsets is considered "long", * aka out of cache. * note : "long" definition seems overloaded here, sometimes meaning "wider than bitstream register", and sometimes meaning "farther than memory cache distance". * This function will try to mitigate main memory latency through the use of prefetching */ static size_t ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, const void* seqStart, size_t seqSize, int nbSeq, const ZSTD_longOffset_e isLongOffset, const int frame) { DEBUGLOG(5, "ZSTD_decompressSequencesLong"); #if DYNAMIC_BMI2 if (ZSTD_DCtx_get_bmi2(dctx)) { return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame); } #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */ #if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \ !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG) /* ZSTD_getLongOffsetsShare() : * condition : offTable must be valid * @return : "share" of long offsets (arbitrarily defined as > (1<<23)) * compared to maximum possible of (1<<OffFSELog) */ static unsigned ZSTD_getLongOffsetsShare(const ZSTD_seqSymbol* offTable) { const void* ptr = offTable; U32 const tableLog = ((const ZSTD_seqSymbol_header*)ptr)[0].tableLog; const ZSTD_seqSymbol* table = offTable + 1; U32 const max = 1 << tableLog; U32 u, total = 0; DEBUGLOG(5, "ZSTD_getLongOffsetsShare: (tableLog=%u)", tableLog); assert(max <= (1 << OffFSELog)); /* max not too large */ for (u=0; u<max; u++) { if (table[u].nbAdditionalBits > 22) total += 1; } assert(tableLog <= OffFSELog); total <<= (OffFSELog - tableLog); /* scale to OffFSELog */ return total; } #endif size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, const int frame, const streaming_operation streaming) { /* blockType == blockCompressed */ const BYTE* ip = (const BYTE*)src; /* isLongOffset must be true if there are long offsets. * Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN. * We don't expect that to be the case in 64-bit mode. * In block mode, window size is not known, so we have to be conservative. * (note: but it could be evaluated from current-lowLimit) */ ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || (dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN)))); DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize); RETURN_ERROR_IF(srcSize >= ZSTD_BLOCKSIZE_MAX, srcSize_wrong, ""); /* Decode literals section */ { size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, streaming); DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize); if (ZSTD_isError(litCSize)) return litCSize; ip += litCSize; srcSize -= litCSize; } /* Build Decoding Tables */ { /* These macros control at build-time which decompressor implementation * we use. If neither is defined, we do some inspection and dispatch at * runtime. */ #if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \ !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG) int usePrefetchDecoder = dctx->ddictIsCold; #endif int nbSeq; size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, srcSize); if (ZSTD_isError(seqHSize)) return seqHSize; ip += seqHSize; srcSize -= seqHSize; RETURN_ERROR_IF(dst == NULL && nbSeq > 0, dstSize_tooSmall, "NULL not handled"); #if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \ !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG) if ( !usePrefetchDecoder && (!frame || (dctx->fParams.windowSize > (1<<24))) && (nbSeq>ADVANCED_SEQS) ) { /* could probably use a larger nbSeq limit */ U32 const shareLongOffsets = ZSTD_getLongOffsetsShare(dctx->OFTptr); U32 const minShare = MEM_64bits() ? 7 : 20; /* heuristic values, correspond to 2.73% and 7.81% */ usePrefetchDecoder = (shareLongOffsets >= minShare); } #endif dctx->ddictIsCold = 0; #if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \ !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG) if (usePrefetchDecoder) #endif #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame); #endif #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG /* else */ if (dctx->litBufferLocation == ZSTD_split) return ZSTD_decompressSequencesSplitLitBuffer(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame); else return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame); #endif } } void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst, size_t dstSize) { if (dst != dctx->previousDstEnd && dstSize > 0) { /* not contiguous */ dctx->dictEnd = dctx->previousDstEnd; dctx->virtualStart = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart)); dctx->prefixStart = dst; dctx->previousDstEnd = dst; } } size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) { size_t dSize; ZSTD_checkContinuity(dctx, dst, dstCapacity); dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 0, not_streaming); dctx->previousDstEnd = (char*)dst + dSize; return dSize; } |